The GBR is a unique data source that contains information collected during ANC, delivery, and the postpartum period for almost all deliveries in Georgia. The results of the present study confirmed that most of the selected antenatal, intrapartum, and newborn variables registered in the GBR were in reasonable agreement with the information registered in the MRs. The exceptions were the number of previous spontaneous abortions, ANC visits, GA week at delivery, and fetal heartbeat upon hospital admission. It is important to keep in mind that this study evaluates the transfer of data from MRs to a digital registry (the GBR) and, not errors in the MRs themselves. In this study, the proportion or errors was mainly below 5%, which is comparable to other studies using single-entry data processing methods [5].
Antenatal variables
Of the selected ANC-related information, the registration of parity and the number of previous CSs in the GBR displayed 96% and 92% complete agreement with the MRs, respectively. There was a larger proportion of missing information regarding previous CSs in the GBR than in the MRs, which explains some of the disagreement between the registrations. Likewise, the number of previous spontaneous abortions had almost twice as many missing values in the GBR as in the MRs, which contributed to a lower proportion of registrations with complete agreement (81%). Interestingly, if all entries with missing information in either MRs or GBRs was excluded, the proportion of complete agreement would increase to 96.5% for previous spontaneous abortions. In 2018, it was not mandatory to register the number of previous spontaneous abortions in the GBR, which explains the larger number of missing values in the GBR entries. However, since 2019, registration has become mandatory; hence, the number of missing values in GBR has decreased considerably. In 2019, the registration of previous CS became mandatory in the GBR. Disagreements in the registration of the number of ANC visits between the GBR and MRs are explained by changes in the reimbursement process related to ANC visits. In Georgia, costs related to ANC are reimbursed by the Social Service Agency, and after the implementation of the GBR, the reimbursement process was conducted electronically. In the present study, only 73% of the registrations were in agreement with the MRs, of which 70% were recorded with a higher number in the GBR than in the MR. Thus, because the GBR is currently the primary data source for reimbursement, it is highly likely that the registration of ANC visits is more complete in the GBR than in the MRs.
Although we demonstrated excellent agreement between the registered information on maternal morbidity in the GBR and the corresponding information in the MRs, our results also highlight that the prevalence of gestational diabetes, preeclampsia, and hypertensive disorders registered in the MRs was low. Only 0.2% of pregnant women in Georgia were registered with gestational diabetes, whereas recent systematic reviews and meta-analyses estimated its prevalence in Europe and Asia to be approximately 11% [6, 7]. Large regional variations in prevalence have been reported; for example, the pooled prevalence in Northern Europe in 2014–2019 was approximately 9% [6], 31.5% in Eastern Europe [6], 1.5% in Nepal [7], and 22.9% in Saudi Arabia [7]. These large differences can be partly explained by the various registration regimes, diagnostic criteria, and testing strategies, as there is no universal consensus. Nevertheless, a prevalence of less than 1% in Georgia is highly unlikely and suggests that either pregnant women remain undiagnosed, which is unfortunate since gestational diabetes increases the risk of delivery by CS, macrosomia, preterm birth, low 1-minute Apgar score, and born large for gestational age [8], or that the disease is poorly registered in the MRs. Likewise, although there are disagreements worldwide regarding the classification and diagnosis of preeclampsia and hypertensive disorders in pregnancy [9], the prevalence of these conditions registered in the MRs in Georgia (preeclampsia:0.3% and hypertensive disorders:0.6%) was considerably lower than the crude global prevalence of preeclampsia of 2.3% in 2002–2010 [10], although regional variations exist, e.g., 4.2% in the western Pacific region and 1.2% in the eastern Mediterranean region. Hypertensive disorders are usually present in 5.2–8.2% of all pregnancies [11], clearly highlighting that the prevalence registered in the MRs in Georgia is suspiciously low. Because hypertensive disorders during pregnancy are a major cause of maternal and newborn morbidity and mortality [12, 13], it is crucial that women are correctly diagnosed with ANC and receive optimal treatment. Hence, Georgian stakeholders should identify whether the low prevalence of preeclampsia and hypertensive disorder in MRs is due to a lack of diagnosis and, accordingly, proper management or if it is solely a registration problem.
Intrapartum variables
More than 96.5% of the registrations of the selected intrapartum-related information in the GBR were in agreement with the information registered in the MRs, which can be considered sufficient. For instance, 99.1% of delivery dates were the same in the GBR and MRs, and disagreeing registrations can be explained by registration mistakes in either the GBR or MRs. Registration of fetal presentation in the GBR had excellent agreement with MRs; however, according to the MRs, approximately 4.0% of fetuses had a breech presentation and 1.5% had a transverse lie. The prevalence of breech presentation varies with gestational age [14] and complicates approximately 3–5% of pregnancies, which is in line with the registrations in MRs [15]. In contrast, transverse lie is a very rare condition that affects less than 0.5% of term pregnancies [15]. In a study of 11,957 singleton deliveries over a 10-year period in Finland, a transverse lie was present in 0.12% of deliveries [16]. Based on these numbers, the prevalence of 1.5% of transverse lies according to the MRs seems high. Transverse lie is an absolute indication for CS delivery, and Georgia has among the highest CS rates in the world, reaching 43.7% in 2021 [4]. The emergency CS rate in Georgia is also unnaturally high, which could indicate the misclassification of planned CS as an emergency CS [17]. The clinical guidelines in Georgia state that CS should only be performed upon medical indication and that obstetricians are not encouraged to perform CS upon maternal request without a medical indication [18]. Therefore, it is surprising that in 24% of the CS deliveries, maternal requests were registered in the MRs as an indication for CS, clearly suggesting that the clinical guidelines were not entirely followed. The healthcare system in Georgia is privatized, and CS births receive a higher monetary reimbursement from the state than vaginal deliveries, which could be a driver of the high CS rates in the country [19]. In the present study, we also found that 2.2% of CS deliveries had transverse lies as an indication for CS registered in the MRs. This number also appears unnaturally high and may be due to misclassification; however, validation of the MRs was outside the scope of this study.
Although PPH registrations in the GBR demonstrated close to 100% agreement with the data in MRs, the registered prevalence in the MRs was only 0.9%. PPH within 24 h of delivery occurs in 1.2–12.5% of deliveries and is a leading cause of maternal morbidity and mortality; hemorrhage after 24 h is much less common and occurs in <1% of deliveries [20]. In Norway, almost 32% of women delivering in 2021 experienced blood loss of 500 mL or more, whereas 4.5% of mothers lost more than 1500 mL and required blood transfusion [21]. In general, misdiagnosis of PPH is common, as indicated by large variations in prevalence, mainly because of the underestimation of blood loss, lack of proper clinical protocols, and lack of education and training for medical personnel [22]. The low registered prevalence in Georgia is a concern, especially because hemorrhage was identified as the leading direct cause of maternal deaths in Georgia from 2014 to 2017 [23]. Hence, it is important to identify the reason for the low reported prevalence of PPH and assess whether clinical practices regarding the diagnosis and treatment of PPH are adequate.
Newborn variables
Of the selected newborn variables, all except fetal heartbeat upon hospital admission displayed reasonable agreement between registrations in the GBR and MRs. GA weeks at delivery were in complete agreement with the MRs in approximately 94% of the registrations. There were 54 women who had two different GA weeks registered in the MRs, which were not registered in the GBR because of programmed edit checks. Because of these reporting mistakes, these registrations were coded as missing in the MRs, which reduced the proportion of observations with complete agreement between the GBR and MRs. There were no substantial differences in newborn birthweights registered in the two data sources, and small disagreements did not vary with newborn birthweight. Thus, it is likely that the small disagreements between the GBR and MRs are random. Almost 10% of participants lacked information about fetal heartbeat upon hospital admission in the GBR, whereas the information was present in the MRs. It is challenging to speculate about the reason for lack of registration of this variable in the GBR, but the variable is described as “monitoring of fetal heartbeat with cardiotocography upon admission to the hospital” in the GBR, which is a very specific phrase. Due to a lack of time, heartbeat monitoring may not have been performed upon admission but could have been performed later and therefore registered in the MRs but not in the GBR. In either case, our results show that variables related to the registration of fetal heartbeats upon hospital admission should be used and interpreted with caution.
More than 99% of the registrations in the GBR regarding newborn diagnoses such as respiratory distress, newborn infection, and congenital malformations were in complete agreement with the MRs. Based on the MRs, 5.0% of newborns were diagnosed with respiratory distress, 2.0% had infections, and 1.2% had congenital malformations. Worldwide, approximately 7% of newborns are diagnosed with respiratory distress [24]; however, the incidence varies across countries and is higher among premature children [25]. Additionally, children delivered via CS usually have a higher incidence of this condition [26]. In Norway, the incidence of respiratory distress was approximately 1.6% in 2021 [21]. Given that both the proportion of CS and the incidence of early neonatal deaths (the main cause of death being prematurity) in Georgia are considerably higher than those in Norway, an incidence of 5% could be reasonable. It is challenging to compare the number of newborn infections with other studies, as it is not expected to be similar across countries, owing to variations in resources and clinical practices. Additionally, infections diagnosed after transfer to the NICU may not be registered in the GBR. Therefore, this variable should be interpreted with caution.
The incidence of 1.2% of congenital malformations registered in the MRs is considerably lower than that in Sweden in 2016 (3.2% of liveborns) [27] and in Norway in 2021 (4.0%) [21]. However, we did not expect a comparable incidence, as the MRs and GBR only covered the period from birth until discharge from the hospital, while both Norway and Sweden had a longer registration period and included malformations from abortions performed after 12 weeks of gestation. This is a limitation of the data in the GBR and, in fact, of many birth registries. Unless data on congenital malformations are complete from early pregnancy until a certain time after delivery (e.g., 1 year), these data should be used with caution.
Finally, the registration of NICU admissions in the GBR displayed close to 100% agreement with the MRs (99.2%), which was reassuring. NICU admission registrations in the GBR have previously been validated by crosschecking registrations with a hospitalization registry in Georgia [28]. During that study, we found that only 0.39% of NICU transfers were not registered in the GBR, which was a very small proportion. Hence, information regarding NICU admissions in the GBR can be considered of high quality.
Strengths and limitations
This study assessed the quality of registration of selected core variables in the GBR compared with the corresponding information in the MRs. The random selection of participants ensured that the results reflected the quality of registration across all types of health facilities in Georgia. However, our study did not validate the correctness of information on MRs. For such a study, a third source of information is needed. Because of the COVID-19 pandemic, the extraction of data for this study was delayed; since 2018, the GBR has implemented more programmed edit checks and made additional core variables mandatory. Therefore, the current quality of registration for several core variables in the GBR is likely higher than that presented in this study.